From mboxrd@z Thu Jan 1 00:00:00 1970 From: George Dunlap Subject: Re: [PATCH 0/7] Optionally add libvchan to stubdom Date: Fri, 26 Apr 2013 15:57:41 +0100 Message-ID: <517A95E5.3080209@eu.citrix.com> References: <517A959D.9060302@invisiblethingslab.com> Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; Format="flowed" Content-Transfer-Encoding: 7bit Return-path: In-Reply-To: <517A959D.9060302@invisiblethingslab.com> List-Unsubscribe: , List-Post: List-Help: List-Subscribe: , Sender: xen-devel-bounces@lists.xen.org Errors-To: xen-devel-bounces@lists.xen.org To: Marek Marczykowski Cc: Stefano Stabellini , Ian Jackson , Ian Campbell , "xen-devel@lists.xen.org" List-Id: xen-devel@lists.xenproject.org On 26/04/13 15:56, Marek Marczykowski wrote: > On 26.04.2013 16:55, George Dunlap wrote: >> On Fri, Apr 26, 2013 at 3:45 PM, Marek Marczykowski >> wrote: >>> This patch series enhances stubdomain with libvchan. It is disabled by default, >>> but can be easily enabled in mini-os.cfg when needed. >>> In order to do so, some parts of minios needs to be extended: >>> - xenstore (to support transactions) >>> - libxc - to support xc_gntshr >>> >>> This series assumes "libxc: fix xc_gntshr_munmap" already applied. >>> >>> One unsolved problem is implementation of xc_gntshr_share_page_notify. On Linux, >>> there is kernel, which can notify remote in case of server process death. In >>> stubdom there is no such place - if exit() is called anywhere, the whole >>> stubdom is terminated. Also the common case is stubdom destruction (the case >>> for ioemu stubdom - it is destroyed by toolstack at domain shutdown), I don't >>> know if stubdom even support clean shutdown. >>> Anyway I can implement some notification (which needs some additional info >>> stored by minios version of gnttab), just for clean libxenvchan_close(). But it >>> is rather rare case. >>> Some workaround can be checking for domain state at client side, but IMHO it >>> isn't the best way (although still used in original Qubes OS vchan version). >>> Any ideas? >>> >>> Patch 2 is independent - should be also useful for normal Linux case. >> I assume your targeting this for 4.4? > Probably yes, I assume it is too late for 4.3, right? Unless you've got a really amazing justification, yes. :-) -George